Blog @ Illumina
Real scientists. Real commentary.

More is More, and Time will Tell- American Scientific Summit Day 1

Amy Cullinan Ph.D
| Jun 09, 2013

On a sultry New Orleans evening, we kicked off the third and final Scientific Summit for 2013 for the Americas region. Some 75 regional attendees sampled the best of the crescent city before a busy day of plenaries and breakout sessions. The first session focus was on the complexities of genetic disease.

Rebecca Crimian, a genetic counselor from the Duke Center for Human Genome Variation (CHGV) walked us through two examples of exome sequencing used to explore undiagnosed genetic disease. Over fifty thousand cases of unexplained disease occur annually in the United States, and CHGV’s methods for sequencing the father, mother, and affected child, deciphers de novo and recessive variants, as well as compound heterozygous causal mutations. Even more compelling was a family who blogged the results of their son’s genetic test that identified mutations in N-glycanase 1. The blog was found by the parents of another child with a similar condition, improving their diagnostic odyssey and informing them about family planning options.

Jon Sebat, from the University of California San Diego spoke next on unraveling the genetics of neuropsychiatric disorders. Historically, early efforts to find causal genes with autism spectrum diseases (ASDs) were not notably successful, although certain de novo copy number variants (CNVs) are seen in about 10% of autism cases compared to about 1% in normal. Certain hotspots, or non-allelic homologous recombinations that are either duplications or depletions, were noted in some ASD cases. Looking inside these genomic regions revealed pathogenic changes that often are reciprocal: schizophrenic behaviors are sometimes opposite autistic ones. Moving beyond CNVs to de novo mutations in exonic regions, a small but informative study of monozygotic twins revealed seven mutation types in five genes correlate with variation found in recent exon studies of ASD. Evidence of “mutation clusters” or areas within individual genomes that are frequently mutated, surprisingly— even within highly-conserved areas was shown. A fascinating model of predicting the mutability landscape of the human genome suggests that mutability is intrinsically coupled with functionality, or in other words, our genome is programmed to mutate to affect traits. Future exploits in complex and evolutionary genomics include expanding to whole genomes and looking at coding and noncoding regions to be able to understand highly conserved mutational hotspots.

David Craig from the Translational Genomics Institute (TGen) spoke next on integrating both RNA and DNA data into clinical decision making. Dr. Craig emphasized the complexity and importance of optimizing their whole-genome and transcriptome sequencing pipeline to ensure that it is clinically relevant, and that it is flexible enough to accommodate complex scenarios. Using examples from pediatric rare diseases and a study of late-stage metastatic cancers, Dr. Craig and others’ work highlighted additional information often clarifies the clinical picture significantly, whether finding a nonsynonymous SNP that caused a splicing change that was not predicted by the standard genome interpretation tools, or looking at epigenetic changes with phased exome sequencing.

Jim Knowles from the University of Southern California gave an update on an incredibly exciting project called BrainSpan, which provides transcriptome, epigenome, and genome information of the developing brain, including anatomical mapping of structures from on samples ranging from 4 weeks post conception to 45 years of age. Seated at the Allen Institute for Brain Science, this highly collaborative effort will create a searchable database to let researchers browse for genome and transcriptome information by gene category, and full-color, high-resolution digital brain atlases accompanied by a systematic, hierarchically organized taxonomy of developing human brain structures. Dr. Knowles also talked about his experience doing single cell RNA analysis using the Smart-Seq protocol after isolating primary and cultured neurons by patch-clamp. There is quite a technical challenge of setting up cleanroom facilities to do this kind of work!

Neil Miller, from Children’s Mercy Hospital in Kansas City talked about the challenges and significant achievements of performing rapid diagnosis of monogenic diseases with Stat-Seq. An impressively rapid pipeline leads to answers for anxious families as soon as newborns present symptoms suggestive of genetic disease. These symptoms are rapidly identified using a novel software called SSAGA (symptom and sign assisted genome analysis). A sample is taken and sequenced on HiSeq 2500 in rapid run mode, completing within 18 hours. After primary alignment and variant detection using ISSAC, another tool called RUNES (rapid understanding of nucleotide variant effect software) is used to annotate the variants. Within 10 minutes of this step, a third program called VIKING (variant integration and knowledge interpretation in genomes) harmonizes the symptom information from SSAGA and the variant annotations from RUNES to interpret the case. The benefits of rapid diagnosis are obvious in that answers come within a critical window of the first few days of life, often informing therapeutic interventions while ruling out diseases and unnecessary treatments. Dr. Miller then presented exciting data on a new CLIA test called TaGSCAN covering the coding region of 514 genes known to cause severe childhood-onset diseases, citing a case in which an individual with cerebellar atrophy resolved their 5-year diagnostic odyssey within 3 weeks, and found actionable results that included a simple vitamin supplement.

Kicking off the talks on the genetic etiology of cancer after lunch, Peter Laird from the USC Epigenome Center talked about bisulfite sequencing to uncover epigenetic marks, chromatin remodeling, and insulators to examine the interplay between the cancer genome and epigenome. Dr. Laird presented examples of three scenarios during cancer progression: when epigenetic control is disrupted, when genetics influence the epigenetic landscape, and when epigenetics influences the cellular genetics. In the first example, regions of focal CpG hypermethylation were found to correlate with polycomb targets, and were prone to abnormal methylation. Conversely, long stretches of hypomethylated bases corresponding to nuclear laminin attachment points were observed in tumor cells. In these cases, the Infinium HM450 methylation array allowed them to look at thousands of samples quickly and efficiently. For the second and third scenarios of interplay between epigenetic and the cancer genome, Dr. Laird gave examples of somatic mutations influencing epigenetics in glioblastoma and renal clear cell cancer, as well as inactivation of a post replication DNA repair mechanism involving MLH1 and a homologous recombination repair defect for BRCA1 silencing. At the conclusion of an amazingly complex presentation, Dr. Laird reminded us that since many of these interactions involved changes that were outside of the primary sequence, that there is hope that some of the epigenetic interactions are reversible with new therapeutics. For more details on the interplay of genetics and epigenetics in cancer, check out the review from Shen and Laird in Cell.

Next up was Charles Perou from the University of North Carolina at Chapel Hill presenting on the therapeutic implications of sequencing and gene expression for breast cancer. Dr. Perou introduced us to the “intrinsic subtypes” of breast cancer based on microarray gene expression patterns, which give accurate survival outcomes (See a quick video with Dr. Lisa Carey from UNC explaining these cancer subtypes here). After switching from microarrays to qPCR, a new test comprising 50 genes and validated primer sets designed to work together on FFPE tissue helped to assign a risk of relapse score and help patients weigh the benefits and risks of certain treatment options. For example, using the intrinsic subtypes prognostic indicator suggest that luminal A type cancers will show a benefit to receiving hormone therapy, but not much from additional chemotherapy, while luminal B type cancers will show a benefit for both hormone and chemotherapy. One of the most devastating subtypes is the so-called basal-like or triple negative tumors (missing the estrogen and progesterone receptors, and HER2 negative) often have the lowest survival rates. mRNA-Seq data equaling some 33 trillion bases from The Cancer Genome Atlas reveals a complex and not entirely clear therapeutic path where a wide range of inhibitors and therapies may provide benefits. Dr. Perou then talked about with these basal-like breast cancers, the need to understand how the primary tumors and distant metastases are related- whether the ability to metastasize is an intrinsic property of the primary tumor or whether the groups of metastases that appear arise independently. He then gave us a glimpse of some fascinating research on testing individual primary tumor xenografts (in immunocompromised mice) and testing responses and survival times.

Then final speaker of the afternoon was Steven Jones of the Michael Smith Genome Sciences Centre, of the British Columbia Cancer Agency on complete genome sequencing and clinical oncology. Dr. Jones started off with a somewhat shocking statistic, that in British Columbia, $185 million dollars are spent annually on cancer drugs, of which a large portion doesn’t benefit the patients. What is the best way to use genomic information to accurately match therapies to the tumor? Dr. Jones presented four different cases from a comprehensive study of advanced incurable cancers by which whole genome sequencing was used to examine somatic mutations including copy number variants, SNVs, and loss of heterozygosity. RNA-Seq was used to examine gene expression levels, and all of this information was used for annotation of known variants, pathway analysis, drug target analysis, and a summary of therapeutic options. Data from the first two cases reveal many changes between the primary tumor and metastatic lesions, and evidence of deleterious events such as amplification of VEGFA were thought to be caused by previous treatments. A dramatic example was detailed in the third case, where squamous cell carcinoma presented in two different body locations and with two entirely different genetic backgrounds in the same patient. Eroltinib therapy targeting an EGFR mutation was tailored to a facial tumor, and the tumor “melted away”. The second lesion on the chest eventually responded as well. In a final case, sequence analysis of another squamous cell carcinoma revealed a homozygous loss of SMARCB1, which was not typical for this type of tumor or for this patient’s age. The tumor turned out to share more genetic characteristics with atypical tetroid rhabdoid tumor (ATRT), and the patient was switched to a completely different treatment.

One of the overall themes from this first day of talks was “more is more”- and that is usually better. Integrating genomic information with epigenomic or transcriptomic data brings clarity and often a different focus to complex conditions such as metabolic or mitochondrial disease or in clinical decision making. Another theme was that “time will tell"- when mutations are important in the span of evolutionary time, as in Dr. Sebat’s work, or how metastases differ from primary tumors, in the case of triple-neg breast cancer, or even how fast you can turn around a critical NICU genetic test using next-generation sequencing.

And speaking of time, this was just day one. The next post will cover the talks from our second day at the American Scientific Summit.